What the numbers say
In 2023, 737,170 people arrived in Australia under the ABS 12-in-16-month residency rule, against 219,080 departures, producing a net overseas migration figure of 518,090.[^2] Strip out the 100,720 returning Australian and New Zealand citizens and you have 636,450 foreign nationals arriving in a single year. Of those, 282,570, or 44.4%, arrived on study visas.[^2]
Net migration has since fallen: 438,000 in 2022, 530,700 in 2023, 330,500 in 2024, 300,800 in 2025.[^3] The 2023 total was a post-border-reopening catch-up and should not be treated as the ongoing baseline. But composition is not a catch-up effect. Students accounted for 51.7% of net migration in 2023, and the skilled category, the one the intake is nominally sized for, came to 14.1% combined, with permanent Skill alone at 5.8%.[^2] That composition held across 2019, 2022 and 2023. It is structural, not a spike.
So the intake is not, in the main, a skilled-worker intake. It is a student intake that feeds into the labour market through onshore visa switching. The Commonwealth's own Rapid Review into the Exploitation of Australia's Visa System, the Nixon Review, documented exactly how that pathway works: non-genuine enrolments, providers trading on the work rights attached to study visas, and a regulatory environment that made onshore switching from study into work straightforward.[^4] The Nixon Review argued the system, not the people in it. That distinction matters, and this piece holds to it.
The mechanism: what LLMs absorb first
Here is the AI argument, stated plainly.
Large-language model deployment does not start with trades. It starts with graduate-entry and entry-level knowledge work: document processing, customer query handling, research assistance, data extraction, basic coding support, report drafting. These are the roles that a new graduate or a student on a post-study work visa competes for. They are also the roles with the shortest remaining runway before LLM-assisted tooling either replaces the headcount or compresses the salary enough to make the role unviable at current cost-of-living levels.
The intake is concentrated in exactly the occupational band that automation is absorbing first. That is not a forecast dressed as a fact. It is an argument about the direction of deployment, and it is the argument that makes the student-intake figure the most important number in this debate. Importing 282,570 people on study visas in a single year, into a pathway that leads to entry-level knowledge work, at the moment LLM deployment is accelerating through that exact layer of the labour market, is a structural mismatch. The shortage those visas are nominally filling is temporary. The population effect is permanent.
The Productivity Commission said it plainly in its inquiry into migrant intake: migration "does not offer a long-term panacea" because "immigrants age too".[^5] The dependency-ratio argument for a large intake assumes the intake stays young and productive indefinitely. It does not. Add automation to that finding and the case for the current settings collapses entirely.
Labor's record is the evidence
Labor set the planning levels that produced the 2023 figures. That is a record, not an opinion. The settings were a deliberate policy choice, made after the border reopened, in response to employer lobbying about labour shortages. The shortages were real. The question is whether the response was calibrated correctly, and the composition data says it was not.
A 14.1% skilled share against a 51.7% student share is not a skilled-worker program. It is a volume program with a skilled-worker label. And the volume came at a cost that is now visible in every housing affordability dataset the register holds: deposit-to-income ratios, rental vacancy rates, household cost pressures against wage growth. The policy produced those outcomes. Arguing the policy and the figures it produced is not a personal attack on anyone. It is what accountability looks like.
Net migration has fallen since the 2023 peak, and that is worth acknowledging. But a fall from 530,700 to 300,800 still leaves the annual intake well above the long-run average, and it leaves the structural composition problem entirely unaddressed. Students are still the dominant stream. The onshore switching pathway documented by the Nixon Review is still open. The automation displacement that makes the student pathway specifically problematic is accelerating, not pausing.
Why the Coalition needs One Nation to hold the line
The Coalition's current position on migration is a change of mind, not a demonstrated conviction. In office, Coalition governments ran intakes that were high by historical standards. Angus Taylor's current position on reducing immigration is credible as a statement of intent. It is not yet credible as a prediction of what happens when the Business Council calls.
That is not an attack on the Liberals or the Nationals. They bring the numbers, the ministerial experience and the institutional relationships that a governing coalition requires. The argument is narrower: neither party has demonstrated, in government, the willingness to cut the intake and hold it there against sustained business-lobby pressure. One Nation has not been in government to demonstrate it either, but it has published a platform that commits to exactly that, and it has no institutional relationship with the lobbies that would push back.[^1]
Barnaby Joyce's move to One Nation in August 2026 is the clearest signal yet that the realignment is underway. Joyce brings regional credibility and ministerial experience. One Nation brings the one thing the Coalition has not shown it can hold: a floor on the intake that does not lift the moment a hospitality association issues a press release.
The schedule, not the slogan
One Nation's immigration platform commits to a year-on-year reduction in net overseas migration.[^1] That is a schedule. It can be modelled, costed and defended on air. "We would reduce it" cannot.
The continuing nature of the reduction is what separates this from the pattern both major parties have already run. Labor reduced the intake from the 2023 peak. The Coalition reduced it in the years after the 2016 peak. Both allowed it to drift back up once the immediate political pressure eased. A schedule that continues down, paced against measured displacement, and held there rather than reversed, is a different commitment. It is the only commitment that addresses the structural argument rather than the headline number.
The automation-displacement pacing is the part of this argument that nobody else is making. If the intake is reduced in line with measured displacement in the occupational bands LLMs are absorbing, the reduction is defensible on economic grounds that are not reducible to a culture argument. It is a productivity argument: raise output per person on the existing population rather than grow headline GDP by adding heads. The distinction between GDP growth and GDP-per-capita growth is in the data. Australia ran a per-capita recession while headline GDP rose. That is what a headcount growth model produces.
The technology policy gap
One Nation's published policy index covers 29 areas.[^6] None of them address AI, automation, data infrastructure or telecommunications. That is a gap, not a gotcha. A party that is about to be a governing force will have to form a technology policy as the data centre build-out accelerates, and the positions it already holds on foreign ownership, energy sovereignty and domestic resource use point toward what that policy should look like.
Cheap domestic power is what makes Australian-owned compute viable. One Nation's energy and resources positions already argue for domestic use of Australia's resource base rather than export at commodity prices. Apply that logic to compute infrastructure and the policy writes itself: Australian-owned data centres, powered by domestic energy at industrial rates, running on hardware procured through transparent Commonwealth processes. The foreign ownership and foreign lobbying arguments that the section holds in reserve for future pieces connect directly to this gap.
The constructive version of this observation is that One Nation has the raw material for a technology policy that is coherent with everything else it argues. The work is to build it before the election rather than after.